AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

GLASS Flows: Efficient Inference for Reward Alignment of Flow and Diffusion Models

Peter Holderrieth, Uriel Singer, Tommi Jaakkola, Ricky T. Q. Chen, Yaron Lipman, Brian Karrer

ICLR 2026 · Image generation › Fast diffusion sampling · approach: Flow matching distillation

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace SDE transitions with GLASS Flows (rho = 0.4) in Feynman-Kac steeringFKS on FLUX · ImageReward, N=8ImageReward: 1.19 -> 1.40Improved
replace SDE transitions with GLASS Flows (rho = 0.4) in FKSFKS on FLUX · ImageReward, N=8GenEval: 63.8 -> 74.3Improved
replace SDE transitions with GLASS Flows (rho = 0.4) in FKSFKS on FLUX · CLIP reward, N=8CLIP reward: 39.0 -> 39.8Improved
replace SDE transitions with GLASS Flows (rho = 0.4) in FKSFKS on FLUX · Pick reward, N=8GenEval: 64.1 -> 72.6Improved
use FKS with SDE instead of Best-of-N with SDESDE sampling · ImageReward, N=8GenEval: 65.3 -> 63.8No clear effect
use ODE instead of SDE for Best-of-NBest-of-N · ImageReward, N=8ImageReward: 1.16 -> 1.31Improved
add SMC resampling (FKS) to GLASS Flows with rho = 0.4 (ablation)GLASS Flows · ImageReward, N=8ImageReward: 1.32 -> 1.40Improved
set GLASS correlation rho = 0.4 instead of DDPM-equivalent rho (ablation)FKS-GLASS · ImageReward, N=8ImageReward: 1.37 -> 1.40Improved
set GLASS correlation rho = 0.4 instead of DDPM-equivalent rho (ablation)FKS-GLASS · ImageReward, N=8GenEval: 68.2 -> 74.3Improved
add reward gradient guidance to FKS-GLASSFKS-GLASS · FLUX 672x672, ImageRewardImageReward: 1.45 -> 1.52Improved
add reward gradient guidance to FKS-GLASSFKS-GLASS · FLUX 672x672, ImageRewardGenEval: 72.7 -> 73.1Improved
use GLASS Flows instead of DDPM for posterior sampling at few simulation stepsposterior sampling · ImageNet256 SiTFIDImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy